Research On Recommendation Algorithm Based On Bipartite Network

With the advent of the era of big data,the problem of information overload has become more and more serious.The recommendation system is an effective way to solve the information overload.It is used to recommend transactions that may be of interest to users.The recommendation algorithm is the core of the recommendation system.It is used to process input information and form it into recommendation information.In recent years,bipartite graph-based recommendation algorithms have attracted the attention of many researchers.This thesis analyzes the advantages and disadvantages of bipartite graph recommendation algorithms and improves them.The main work is as follows:1.Aiming at the problem of the unreasonable initial resource setting in the traditional bipartite graph recommendation algorithm and the adjustment of resource allocation coefficients based solely on the degree of projects and users,research a bipartite graph recommendation algorithm based on differentiated resource allocation.First,the method of scoring normalization and the maximum and minimum values were used to modify the initial resources of the project.Based on this,the Ebbinghaus forgetting function was used to quantify the impact of the user's "interest shift." Then use the user score similarity function and the user preference function to set the allocation coefficients of the two stages to make the resource flow more reasonable.Finally,experimental verification is performed on the classic Movie Lens dataset and Netflix dataset of the recommendation system.The results show that the newly proposed algorithm improves the accuracy and recall rate while improving the diversity of recommendations.2.Aiming at the problems of user cold start in bipartite graph recommendation algorithm,a bipartite graph recommendation algorithm based on Markov chain trust network is researched.First,the resource value of the project flow to the user during the first stage of the resource allocation process of the bipartite graph recommendation algorithm based on differentiated resource allocation is used as the implicit trust value of the target user to other users,and the direct trust value is obtained by combining the explicit trust value;then The transfer process of the trust value is regarded as a Markov chain model.The Markov chain is used to propagate and aggregate the direct trust value,and the small world theory and trust entropy are used to set the termination condition of the iteration.Finally,the global trust obtained after the iteration is used.Generate recommendations for target users.Experimental verification in the Epinions data set with trust relationships between users shows that the new algorithm not only improves the recommended user coverage,but also the accuracy of the recommendation.